Absolute Configurations and Chitinase Inhibitions of Quinazoline-Containing Diketopiperazines from the Marine-Derived Fungus Penicillium polonicum

Three new quinazoline-containing diketopiperazines, polonimides A–C (1–3), along with four analogues (4–7), were obtained from the marine-derived fungus Penicillium polonicum. Among them, 2 and 4, 3 and 5 were epimers, respectively, resulting the difficulty in the determination of their configurations. The configurations of 1–3 were determined by 1D nuclear overhauser effect (NOE), Marfey and electron circular dichroism (ECD) methods. Nuclear magnetic resonance (NMR) calculation with the combination of DP4plus probability method was used to distinguish the absolute configurations of C-3 in 3 and 5. All of 1–7 were tested for their chitinase inhibitory activity against OfHex1 and OfChi-h and cytotoxicity against A549, HGC-27 and UMUC-3 cell lines. Compounds 1–7 exhibited weak activity towards OfHex1 and strong activity towards OfChi-h at a concentration of 10.0 μM, with the inhibition rates of 0.7%–10.3% and 79.1%–95.4%, respectively. Interestingly, 1–7 showed low cytotoxicity against A549, HGC-27 and UMUC-3 cell lines, suggesting that good prospect of this cluster of metabolites for drug discovery.


Introduction
The biosynthetic gene clusters in one fungus species usually destine the generation of structurally versatile secondary metabolites [1]. In recent years, plenty of structurally-unique secondary metabolites have been obtained from the marine-derived fungi [2,3]. However, natural compounds from fungi usually have a high degree of chiral variety, forming one structure with different configurations, which caused a challenging task of determination for their configurations, especially when the molecules displayed high conformational flexibility [3]. Meanwhile, the stereochemistry of molecules has become one of the most important features of chiral natural products, which play a fundamental role in biology, chemistry and medicine. Thus, assigning the stereochemical characterizations attracted more and more attentions in the field of natural medicinal chemistry [4,5]. In our research on structurally-unique and biologically-active metabolites from the marine-derived fungi, the fungal strain Penicillium polonicum HBU-114, whose EtOAc extract exhibited the original thin-layer chromatography (TLC) and high performance liquid chromatography-under voltage (HPLC-UV) profiles of the secondary metabolites, differed from the other fungi, attracted our attention. HPLC-guided separation resulted in the isolation of three new quinazoline alkaloids, polonimides A-C (1-3) and four known analogues, aurantiomide quinazoline alkaloids, polonimides A−C (1−3) and four known analogues, aurantiomide C (4) [6], anacine (5) [6], aurantiomide A (6) ( Figure S24) [6] and aurantiomide B (7) ( Figure S25) [6] (Figure 1). Herein, we report the isolation, absolute configurations and bioactivity of these compounds.
With careful inspection and analyses of the 1D-NMR and HSQC data ( Figure S3), it was found that 1 shares the same quinazoline core as aurantiomide C (4), a diketopiperazine alkaloid isolated from the fungus Penicillium aurantiogriseum [6]. The main differences between 1 and 4 were the presence of an additional methoxy group (δH 3.42; δC 51.3) and the absence of two nitrogen-hydrogen proton signals (δH 7.27 and 6.73 in 4) ( Figure S22) in 1, suggesting the presence of 17-OCH3 group in 1 instead of 17-NH2 group in 4. The above deduction was further confirmed by the key HMBC correlation from -OCH3 to C-17 of 1 (Figure 2, Figure S5). Polonimide B (2) was also obtained as a yellow amorphous powder with the molecular formula C18H20N4O3 (11 degrees of unsaturation) by HRESIMS spectrum. Detailed analysis of the 1D and 2D NMR spectra of 2 ( Figure S7-S11), it was found that 2 was an analogue of 1. Combination with their NMR (Tables 1 and 2) and HRESIMS data ( Figures S6 and S13), showed that 2 differed from 1 by loss of an OCH3 unit (δH 3.42, δC 51.3 in 1) and replaced by a NH2 unit (δH 7.26 and 6.71 in 2) in 2. Combined analysis of the differences existed in the chemical shifts of H-18, H-19, C-18 and C-19 between 2 and 4 indicated that 2 was an isomer of 4 with different geometries of the double bone C3=C18.

Results
Polonimide A (1) was obtained as a yellow amorphous powder. The molecular formula of C 19 H 21 N 3 O 4 (11 degrees of unsaturation) was established by the positive high resolution electrospray ionization mass spectroscopy (HRESIMS) data. The downfield of 1 H NMR spectrum (Table 1, Figure S1) for 1 exhibited six proton signals, including one nitrogen-hydrogen proton signal δ H 10.49 and five olefin proton signals (δ H 8.13, 7.84, 7.69, 7.52 and 6.22). In addition, the 1 H NMR spectrum of 1 showed one methoxy singlet δ H 3.42 and two methyl doublets (δ H 1.08 and 1.05). The 13 C NMR spectrum (Table 2, Figure S2) of 1 revealed three carbonyl carbon signals (δ C 159.9, 165.2 and 171.8). With careful inspection and analyses of the 1D-NMR and HSQC data ( Figure S3), it was found that 1 shares the same quinazoline core as aurantiomide C (4), a diketopiperazine alkaloid isolated from the fungus Penicillium aurantiogriseum [6]. The main differences between 1 and 4 were the presence of an additional methoxy group (δ H 3.42; δ C 51.3) and the absence of two nitrogen-hydrogen proton signals (δ H 7.27 and 6.73 in 4) ( Figure S22) in 1, suggesting the presence of 17-OCH 3 group in 1 instead of 17-NH 2 group in 4. The above deduction was further confirmed by the key HMBC correlation from -OCH 3 to C-17 of 1 (Figure 2, Figure S5).
With careful inspection and analyses of the 1D-NMR and HSQC data ( Figure S3), it was found that 1 shares the same quinazoline core as aurantiomide C (4), a diketopiperazine alkaloid isolated from the fungus Penicillium aurantiogriseum [6]. The main differences between 1 and 4 were the presence of an additional methoxy group (δH 3.42; δC 51.3) and the absence of two nitrogen-hydrogen proton signals (δH 7.27 and 6.73 in 4) ( Figure S22) in 1, suggesting the presence of 17-OCH3 group in 1 instead of 17-NH2 group in 4. The above deduction was further confirmed by the key HMBC correlation from -OCH3 to C-17 of 1 ( Figure 2, Figure S5). Polonimide B (2) was also obtained as a yellow amorphous powder with the molecular formula C18H20N4O3 (11 degrees of unsaturation) by HRESIMS spectrum. Detailed analysis of the 1D and 2D NMR spectra of 2 ( Figure S7-S11), it was found that 2 was an analogue of 1. Combination with their NMR (Tables 1 and 2) and HRESIMS data ( Figures S6 and S13), showed that 2 differed from 1 by loss of an OCH3 unit (δH 3.42, δC 51.3 in 1) and replaced by a NH2 unit (δH 7.26 and 6.71 in 2) in 2. Combined analysis of the differences existed in the chemical shifts of H-18, H-19, C-18 and C-19 between 2 and 4 indicated that 2 was an isomer of 4 with different geometries of the double bone C3=C18. Polonimide B (2) was also obtained as a yellow amorphous powder with the molecular formula C 18 H 20 N 4 O 3 (11 degrees of unsaturation) by HRESIMS spectrum. Detailed analysis of the 1D and 2D NMR spectra of 2 (Figures S7-S11), it was found that 2 was an analogue of 1. Combination with their NMR (Tables 1 and 2) and HRESIMS data ( Figures S6 and S13), showed that 2 differed from 1 by loss of an OCH 3 unit (δ H 3.42, δ C 51.3 in 1) and replaced by a NH 2 unit (δ H 7.26 and 6.71 in 2) in 2. Combined analysis of the differences existed in the chemical shifts of H-18, H-19, C-18 and C-19 between 2 and 4 indicated that 2 was an isomer of 4 with different geometries of the double bone C 3 =C 18 .
Polonimide C (3) was obtained with the molecular formula of C 18 H 22 N 4 O 3 (10 degrees of unsaturation) by the positive HRESIMS data ( Figure S21). Analysis of the 1 H and 13 C NMR spectra of 3 (Tables 1 and 2, Figures S14-S17), revealed 3 had the same anacine core structure as 2. The main differences between 3 and 2 were the NMR chemical shifts of 2-NH, C-4, C-19 and the presence of one nitrogen-bearing methine (δ H 4.74, δ C 55.7 in 3), one methylene (δ H 2.29/1.66, δ C 39.0 in 3) and the absence of one olefinic quaternary carbon (C-3 in 2) and methylene (C-18 in 2) in 3. It was inferred that the double bone between C-3 and C-18 in 2 was reduced in 3, which could also be verified by the 1 Figure S18). The chemical structure of 3 was further confirmed by the key HMBC correlations from NH-2 to C-3, C-4 and C-14, from H-14 to C-1 and C-4, from H-18 to C-4 and from H-3 to C-4 of 3 ( Figure 2, Figure S19). It should be noted that the chemical structure of 3 could be found in the Scifinder database with the cas registry number 154725-83-4. However, the chemical structure of cas 154725-83-4 was the same as anacine (5), which was originally proposed as a benzodiazepine structure by Mantle and co-workers [7] but was revised as a quinazoline structure by Sim and co-workers [8].
The geometries of the double bond C 3 =C 18 in 1 and 2 were established with their selective 1D nuclear overhauser effect (NOE) experiments ( Figure S12). In compound 1, the irradiation of 2-NH (δ H 10.49) resulting in no obvious enhancement of H-18 (δ H 6.22) indicated that 2-NH and H-18 might be trans oriented; the downfield chemical shift of H-18 caused by the deshielding effect of 4-imine also confirmed the Z geometries of the double bond C 3 =C 18 . Whereas, in compound 2, irradiation of 2-NH (δ H 10.49) resulted in enhancement of H-18 (δ H 5.54), was ascertained that H-18 and 2-NH in 2 were cis oriented, which indicated the geometries of the double bond C 3 =C 18 was E.
The absolute configuration of the Glutamine residue at C-14 position of 1-3 was established by the combination of Marfey's method and comparing the computed electronic circular dichroism (ECD) spectra with their experimental results. The HPLC analysis showed that the Glutamine residue in compounds 1-3 was L-Glutamine ( Figure 3). Simultaneously, the ECD curve displayed that the predicted ECD spectra of 14S-1/2/3 look similar to the experimental results of 1-3 ( Figure 4, Figures S26-S31). Therefore, the absolute configuration of C-14 was assigned as S in 1-3. The geometries of the double bond C3=C18 in 1 and 2 were established with their selective 1D nuclear overhauser effect (NOE) experiments ( Figure S12). In compound 1, the irradiation of 2-NH (δH 10.49) resulting in no obvious enhancement of H-18 (δH 6.22) indicated that 2-NH and H-18 might be trans oriented; the downfield chemical shift of H-18 caused by the deshielding effect of 4-imine also confirmed the Z geometries of the double bond C3=C18. Whereas, in compound 2, irradiation of 2-NH (δH 10.49) resulted in enhancement of H-18 (δH 5.54), was ascertained that H-18 and 2-NH in 2 were cis oriented, which indicated the geometries of the double bond C3=C18 was E.
In    Analysis of the structure of 3 and the experimental ECD curves of 1−3 (Figure 4) revealed the absolute configuration of C-3 in 3 contributed little to its ECD Cotton effects. Since the absolute configuration at C-14 has been determined by Marfey's analysis, the 13 C NMR chemical shift calculation was applied to confirm the C-3 absolute configuration of 3, with the combination of DP4plus probability method, which is one of the most sophisticated and popular strategies for chemical structure interpretation [9]. Both 3 and 5 were performed with two configurations Analysis of the structure of 3 and the experimental ECD curves of 1-3 (Figure 4) revealed the absolute configuration of C-3 in 3 contributed little to its ECD Cotton effects. Since the absolute configuration at C-14 has been determined by Marfey's analysis, the 13 C NMR chemical shift calculation was applied to confirm the C-3 absolute configuration of 3, with the combination of DP4plus probability method, which is one of the most sophisticated and popular strategies for chemical structure interpretation [9]. Both 3 and 5 were performed with two configurations [(3R,14S)-3/5 and (3S,14S)-3/5)] for NMR calculations. The calculated NMR data of 3 and 5 were all together compared with the experimental results, respectively. The result showed that (3R,14S)-3 was more likely than (3S,14S)-3 (100% vs. 0%) compared with the experimental data of 3 and (3S,14S)-5 is more likely than (3R,14S)-5 (100% vs. 0%) compared with the experimental data of 5 ( Figures S32 and S33). Subsequently, the absolute configurations of 3 and 5 were assigned as 3R,14S-3 and 3S,14S-5, respectively.
The insect enzymes GH20 β-N-acetyl-d-hexosaminidase Of Hex1 and GH18 chitinase Of Chi-h represent important chitinolytic enzymes found in the agricultural pest Ostrinia furnacalis (Guenée) and inhibition of these enzymes have been considered a promising strategy for the development of eco-friendly pesticides. All of the isolated compounds were evaluated for their in vitro inhibitory potency against Of Hex1 and Of Chi-h, by using MU-GlcNAc and MU-(GlcNAc) 2 as substrates, respectively. Compounds 1-7 exhibited weak activities towards Of Hex1 and strong activities towards Of Chi-h at a concentration of 10.0 µM (Table 3), suggesting some valuable clues regarding the structure-activity relationships. Compound 1 bearing the methoxy group on C-17 could weaken the inhibitory activities against Of Hex1 and Of Chi-h compared to 4. Moreover, in compounds 2 and 4, the double bond of Z configuration of 4 showed better activities, indicated that the geometries of the double bone C 3 =C 18 in 2 and 4 had a direct influence on the inhibitory efficiency against Of Hex1 and Of Chi-h. To further explore the inhibition mechanism of these quinazoline-containing diketopiperazines towards Of Hex1 and Of Chi-h, compounds 1 and 4 were firstly selected for investigating the binding mode using molecular docking to Of Chi-h ( Figure 5a) and Of Hex1 (Figure 5c) To further explore the inhibition mechanism of these quinazoline-containing diketopiperazines towards OfHex1 and OfChi-h, compounds 1 and 4 were firstly selected for investigating the binding mode using molecular docking to OfChi-h ( Figure 5a) and OfHex1 (Figure 5c), respectively.
Compound 1 was found to tightly bind to the entire active pocket of OfChi-h ( Figure S34). Three hydrogen bonds, with bond lengths of 2.67 Å, 2.82 Å and 2.91 Å, were formed by the C-12 cycloamide-carbonyl group and C-17 ester carbonyl group of 1 with the guanidine group of ARG439 ( Figure S35). The benzene ring and the lactam ring of 1 had an π-sulfur interaction with the sulfur atom of methionine MET381, with the operating distances of 4.57 Å and 5.01 Å, respectively ( Figure  S36). It was also found that 1 had a π-π stacking with the benzene rings of Trp268 and Phe309 ( Figure S37). Alkyl hydrophobic interactions were showed between the C-18 isopropyl of 1 and Ala355/Met381 and between the 17-OCH3 of 1 and Val469 ( Figure S38). In addition, the C-18 isopropyl of 1 bound to Tyr156 and Phe184 through mixed π/alkyl hydrophobic interactions ( Figure  S39).
Compound 4 was found to bind to the OfHex1 in a "U" conformation ( Figure S40). The C-17 carbonyl and 17-NH2 of 4 formed hydrogen bonds on the guanidine group of ARG220 and the carboxyl of ASP367, respectively ( Figure S41). Compound 4 had a π-anion with the carboxyhydroxyl oxygen anion in the residue of GLU368 ( Figure S42) and π-π stacking interaction with the indole ring of Trp490 ( Figure S43). It was also found that the C-18 isopropyl group of 4 had an alkyl hydrophobic interaction with the isopropyl group of Val484 ( Figure S44). Mixed π/alkyl hydrophobic interactions were also found between the C-18 isopropyl group of 4 and Trp322/Trp483 and between the benzene and lactam rings of 4 and the isopropyl group of Val327 ( Figure S45).
It is noteworthy that the geometric isomers 2 and 4 had different activity value, urging to investigate the binding mode of 2 and OfHex1 (Figure 5b). From the FlexibleDocking results, it was found that the π-anion between 4 and the carboxyhydroxyl oxygen anion in GLU368 was absent in 2. Moreover, the amocarbonyl group of 4 and guanidine group of Arg220 formed two hydrogen bond interactions, while there was only one N-H hydrogen bond interaction between 2 and Arg220 ( Figure S46). The above results could explain 2 was less active than 4.  (Table S1). Among them, only compound 5 exhibited cytotoxicities against the three cell lines (IC50 = 6.0, 6.2, 7.2 μM, respectively). Compound 1 was found to tightly bind to the entire active pocket of Of Chi-h ( Figure S34). Three hydrogen bonds, with bond lengths of 2.67 Å, 2.82 Å and 2.91 Å, were formed by the C-12 cycloamide-carbonyl group and C-17 ester carbonyl group of 1 with the guanidine group of ARG439 ( Figure S35). The benzene ring and the lactam ring of 1 had an π-sulfur interaction with the sulfur atom of methionine MET381, with the operating distances of 4.57 Å and 5.01 Å, respectively ( Figure S36). It was also found that 1 had a π-π stacking with the benzene rings of Trp268 and Phe309 ( Figure S37). Alkyl hydrophobic interactions were showed between the C-18 isopropyl of 1 and Ala355/Met381 and between the 17-OCH 3 of 1 and Val469 ( Figure S38). In addition, the C-18 isopropyl of 1 bound to Tyr156 and Phe184 through mixed π/alkyl hydrophobic interactions ( Figure S39).
Compound 4 was found to bind to the Of Hex1 in a "U" conformation ( Figure S40). The C-17 carbonyl and 17-NH 2 of 4 formed hydrogen bonds on the guanidine group of ARG220 and the carboxyl of ASP367, respectively ( Figure S41). Compound 4 had a π-anion with the carboxyhydroxyl oxygen anion in the residue of GLU368 ( Figure S42) and π-π stacking interaction with the indole ring of Trp490 ( Figure S43). It was also found that the C-18 isopropyl group of 4 had an alkyl hydrophobic interaction with the isopropyl group of Val484 ( Figure S44). Mixed π/alkyl hydrophobic interactions were also found between the C-18 isopropyl group of 4 and Trp322/Trp483 and between the benzene and lactam rings of 4 and the isopropyl group of Val327 ( Figure S45).
It is noteworthy that the geometric isomers 2 and 4 had different activity value, urging to investigate the binding mode of 2 and Of Hex1 (Figure 5b). From the FlexibleDocking results, it was found that the π-anion between 4 and the carboxyhydroxyl oxygen anion in GLU368 was absent in 2. Moreover, the amocarbonyl group of 4 and guanidine group of Arg220 formed two hydrogen bond interactions, while there was only one N-H hydrogen bond interaction between 2 and Arg220 ( Figure S46). The above results could explain 2 was less active than 4.

Discussion
Diketopiperazines, which have been found to occur from a wide range of fungi, display a variety of bioactivities from antineoplastic, antifungal, antibacterial, to anti-inflammatory effects and have the potential to be used in the development of new drugs [10,11]. Among them, quinazoline-containing diketopiperazines, generally possess a tricyclic core of benzene−pyrimidinone−diketopiperazine, are relatively rare. Due to the configurational flexibility of the residue and little contribution to ECD Cotton effects, it was hard to assign the absolute configurations of C-3 of 3 and 5. In the course of our study, the 13 C NMR chemical shift calculation with the combination of DP4plus probability method was applied to assign and distinguish the C-3 absolute configuration of 3 and 5. Furthermore, the isomers 2 and 4 and 3 and 5 showed different activity value, suggesting that the geometries of the double bone C 3 =C 18 in 2 and 4 and the configuration at C-3 in 3 and 5, may play an important role for bioactivities.

General Experimental Procedures
Optical rotatory (OR) and ECD data were performed on a JASCO P-1020 and JASCO J-815 spectrometers (JASCO Corporation, Tokyo, Japan), respectively. UV and IR spectra were gathered using a Perkin-Elmer model 241 spectrophotometer (Perkin-Elmer Corporation, MA, USA) and a Nicolet NEXUS 470 spectrophotometer (Thermo Corporation, MA, USA), respectively. 1D/2D NMR spectra were measured on a Bruker AV-600 spectrometer (Thermo Corporation, Karlsruhe, Germany). HRESIMS data were recorded on a Thermo Scientific LTQ Orbitrap XL spectrometer (Thermo Corporation, MA, USA). Semipreparation HPLC (Shimadzu LC-20AT system) (Hitachi High Technologies, Tokyo, Japan) was operated using a SPD-M20A detector (Hitachi High Technologies, Tokyo, Japan) and a Waters RP-18 column (Waters Corporation, Manchester, UK). The material of Sephadex LH-20 and Silica gel used for chromatographic separation were the same as those in our previous literature [4].

General Computational Procedure
Quantum chemical calculations for 1-3 and 5 were carried out on the basis of previous references (gas phase) [12][13][14]. Chemical structures of 1-3 and 5 were constructed and used for conformational searches using MMFF94S force field by the BARISTA 7.0 software (CONFLEX Corporation Tokyo, Japan). Of all the geometries, those with relative energy of 0-10.0 kcal/mol (93 stable conformers for 1, 85 stable conformers for 2, 49 stable conformers for 3 and 58 stable conformers for 5) were optimized at the B3LYP/6-311+G(d) level, then those with a relative energy of 0-2.5 kcal/mol (15 conformers for 1, 27 conformers for 2, 9 conformers for 3 and 16 conformers for 5) were chosen for ECD calculations at the B3LYP/6-311++G(2d,p) level and simulated using SpecDis 1.71 [15]. In addition, DP4plus applications were used to assign the absolute configurations of 3 and 5, the optimized conformers were calculated at the B3LYP/6-311+G(d,p)//B3LYP/6-311+G(d,p) level for the unshielded tensor values. All of the calculations were processed with Gaussian 09 package [16].

Preparation and Analysis of Marfey's Derivatives
Compounds 1-3 (0.2 mg, respectively), dissolved in 0.5 mL of 6N HCl under the temperature of 110 • C, were hydrolyzed for 4 h. After temperature of the solutions dropped to 25 • C, the mixture were evaporated to dryness under vacuum with addition of distilled H 2 O to remove the trace HCl, then redissolved it in H 2 O (50 µL). The divided hydrolysate were treated with 200 µL of 0.5% (w/v) 1-fluoro-2-4-dinitrophenyl-5-L-alanine amide (FDAA) in acetone and 20 µL 1N NaHCO 3 in order of precedence. The mixture was stirred at 45 • C for 40 min and then it was quenched through the addition of 20 µL of 2N HCl. The mixture was evaporated to give the resulting residues, which was dissolved in MeOH (20 µL) [17]. Similarly, the standard amino acid L -Glutamine and D -Glutamine were derivatized with FDAA using the same procedure as that of 1-3. And the derivatives were analyzed by HPLC with linear isocratic elution (MeOH-H 2 O (70:30, v:v)) detected at 254 nm.

Enzymes Inhibitory Activity Assay
In a final assay volume of 100 µL, enzyme was incubated with substrate (20 µM MU-(GlcNAc) 2 for Of Chi-h and 50µM MU-GlcNAc for Of Hex1) in 20 mM sodium phosphate buffer (pH 6.0 for Of Chi-h, pH 6.5 for Of Hex1) containing 10 µM inhibitor at 30 • C. The reaction in the absence of inhibitor was used as a control. After reacting for an appropriate time (30min), an equal volume of 0.5 M Na 2 CO 3 was added to the reaction mixture to terminate the reaction and the fluorescence of the liberated MU was quantitated using a Varioskan Flash microplate reader, with excitation and emission wavelengths of 360 and 450 nm [18].

Molecular Docking
The complex crystal structure of Of Hex1-PUGNAc (Protein Data Bank (PDB) entry code: 3OZP) [19] or Of Chi-h-chitohepatose (PDB entry code: 5GQB) [20] was used as the starting model for molecular docking employing the Discovery Studio 2017 software. Before docking calculations, conformational searches of the compounds 1 and 4 were performed with the GMMX conformer calculation (Force field: MMFF94, energy window: 5.0 kcal/mol) in GaussView 6.0. Then, the top 5 conformations with the lowest energy were optimized at the B3LYP/6-31G(d) basis set using density functional theory (DFT) in Gaussian 16. To simulate real conditions, the solvent effects of H 2 O were studied using the solvation model based on density (SMD). For the protein, the protein preparation processes were undergone, such as removing water molecules, adding hydrogen atoms and supplementing amino acid residues. The Flexible Docking protocol, which allows for some receptor flexibility during docking of flexible ligands [21] that employs CHARMm in Discovery Studio 2017 software, was used in this study. The receptor binding sites were determined from the PDB site records.

Cytotoxic Assay
All of the isolated compounds 1-7 were evaluated for cytotoxic activity in vitro according to MTT method [22]. Three human tumor cell lines were included-human lung cancer cell line (A549), human gastric cancer cell line (HGC-27), human bladder cancer cell line (UMUC-3) and a non-tumoral cell line, human gastric epithelium (GES-1). The positive control was cisplatinum (DDP).
Cell lines and cell culture Human A549 and HGC-27 cancer cells were obtained from Chinese Academy of Medical Sciences, Basic Medicine Cell Center (Beijing, China). Human UMUC-3 cancer cells were obtained from Cell Resource Center, Shanghai Institute of Life Sciences, Chinese Academy of Sciences (Shanghai, China). Cells were cultured in RPMI-1640 media supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, and 100 mg/mL streptomycin at 37 • C in a 5% CO 2 atmosphere.

Conclusions
Three new quinazoline alkaloids polonimides A-C (1-3), were obtained from the marine-derived fungus Penicillium polonicum. The relative and absolute configurations of 1-3 were comprehensively determined by combination of 1D NOE experiments, modified Marfey's analysis, ECD and NMR chemical shift calculations. The quinazoline-containing diketopiperazines (1-7) with low cytotoxicity but potent chitinase inhibitory activity also indicated good prospect of this cluster of metabolites for drug discovery.

Conflicts of Interest:
The authors declare no conflict of interest.